Table 2.

Mapping criteria for identifying cellular objects to algorithmic steps of automated segmentation.

Criterion Type	Criterion Description	Segmentation Algorithmic Step	Abbreviation	Comment
Imaging	Signal from a cell is higher than background noise	Intensity thresholding	T	All voxels with intensity above the chosen threshold become foreground, all other pixels become background
Imaging	A cell touching the edge of the field of view will be discarded	Removal of objects touching the image edges	E	The shape of a cell touching the edge of the field of view cannot be determined since it is cut-off (part of the cell body is outside the field of view)
Geometry	A cell does not contain any enclosed cavities	Spatial filling of cavities	F	Generally, cells are not expected to have cavities within their volume. However, it is conceivable that a cell could have a tunnel if it were wrapped around a fiber, or a void volume if it was wrapped around a spherical object.
Geometry	a) A cell shape is continuous and does not have disconnected parts b) The cell will be the largest object in an image (background debris in the image are smaller than the cell)	Spatial- and intensity-based removal of small objects	L	a) Only one object should remain after segmentation b) The largest object in the image that is not touching the edge of the image will also be the one object remaining after segmentation
Geometry	Lessen contribution of image features below 1 μm in size	Surface smoothing of objects to remove features < 1 μm in size: closing, opening with 3×3×3 kernel that corresponds to 0.72 μm (x) by 0.72 μm (y) by 2.139 μm (z)	M	Although cells have sub-micrometer features, the uncertainty in image data at this size-scale is not reliable, could arise from noise or debris, and may be artifactual